Affine.Earth Math Court — public MCP bind

Deterministic, no-float Math Court for agent builders. Live remote: affine.earth MCP. Glama (approved): Affine Earth Math Court Remote.

Public bind tree: clients/math-court-mcp/ · MATH-COURT-MCP.md

Registry: earth.affine/math-court. Floats refused. Integers sealed.

---

☀️🌑 Eclipse 2026 — the sky writes geometry, and geometry cannot lie

RESULT — first look, T+1 (2026-08-13): the raw receiver data shows a measured TEC depletion at every sealed totality station in custody, inside its sealed window, in the sealed order — Reykjavik at its sealed maximum, the Spanish trio inside the pre-registered lag gate, depths inside the pre-registered band. Sealed 27 days ahead; measured the day after. The pinned-archive verdict lands ~2026-08-30. Full table: the prediction registry. One entry in the track record — no more, no less.

On August 12, 2026 at 17:45:54 UT, the Moon's shadow — a cone almost 380,000 km long with a tip 294 km wide — dragged across the Earth at supersonic speed, over western Iceland and northern Spain. This repository is the complete, public, falsifiable record of an experiment that reads that moment as a three-dimensional shape, not a number — its claims sealed 27 days in advance, its data drawn only from public archives, its verdicts computed in exact integers. No code lives here; this is the experiment's own surface, and its data tables are generated by the substrate from its own ledger.

The realization

Sixty to a thousand kilometers up is a second sky — the ionosphere, the charged shell GPS travels through and aviation radio bounces off. When the shadow cone pierces it, the shell answers with a moving dent: charge collapses 20–60% under the footprint and refills behind it. For sixty years, science measured that dent with a ruler: "TEC dropped 40%." But rulers can be fooled — on May 11, 2024 a geomagnetic superstorm dug deeper dents than any eclipse, at every station, under an unobscured Sun. Every ruler-pipeline running that day recorded an eclipse that never happened.

The substrate reads the dent as shape: the response either threads the shadow's own timetable — confined to the exact minutes the geometry names, deepest within seconds of each station's own maximum, marching station-to-station at the cone's ground speed — or it does not. Linked or not linked: an integer, no maybe. A storm can dig a deeper hole; it cannot thread the ring. It has no cone, no track, no clock.

Because shape cannot be counterfeited and shape-verdicts are exact, one machine sealed to-the-second claims about the sky 27 days before the sky moved — including a verdict that stays decidable through a superstorm — and anyone on Earth can check them against the same archives: the raw first look matched at T+1, and the pinned-archive verdict lands ~2026-08-30.

Read in this order

Page	What it holds
Overview — the realization	The second sky, shape vs ruler, everything sealed
Science research	The 2026 eclipse, measured shadow physics, the reference storms, the solar-cycle noise floor
Historical corpus	Three eclipses, three solar regimes, one frozen law — the band history writes for 2026
Model shear	The storm data that breaks ruler-science — measured — and the discriminant it taught
Blind spots — five stories	What number-science will always miss and shape-science reads plainly
Data archives	Every public source, verified access mechanics, provenance checksums
Substrate architecture	How a shadow becomes a geometry a machine can hold
Benchmark results	The sealed 2024 verdict + cold-machine reproduction runbook
Prediction registry	Both sealed claims, verbatim, with falsification conditions
Satellite & aviation advisory	The only scheduled ionospheric disturbance in history — GNSS, HF, satellite windows
Success criteria	S1–S5 decidable pass/fail, the storm contingency, failure semantics
Operations runbook	Eclipse day in UT, the live loop, the T+1 resolution
The first 27 days — full review	T+1: every failure itemized, every promise graded, the fair test verified

How this stays true

The generating substrate lives elsewhere; after every ingest or verdict it regenerates the marked blocks in these pages from its ledger, and a drift gate proves the wiki matches the data byte-for-byte. The prose frames; the substrate speaks its own numbers.

Resolved on schedule — graded on the archive's clock. The sky moved 2026-08-12; the raw first look matched at every custody station at T+1; the pinned-archive verdict, pass or fail, lands ~2026-08-30 in these same tables. The historical record already publishes its own falsifiers, because a recorded failure is how this experiment learned to see.

---

One method, many skies

The eclipse was the first sky to sign its name. Seven more charters are now sealed and open: rocket burns that punch ~900 km holes in the ionosphere on a manifest clock, solar flares that step the daylit hemisphere in lockstep with the GOES X-ray record, tsunamis that keep a Huygens travel-time appointment at every tide gauge, cosmic-ray dropouts that strobe ~50 neutron monitors from a single L1 shock minute, explosions whose seismic arrival ladder no earthquake can climb, the galactic-center VLBI table for Sagittarius A* — where Earth-rotation tracks and closure invariants grade what the Milky Way actually wrote before any ring image fills the gaps — and the astrometric-spectroscopic orbit of Gaia BH1, where a star's measured wobble carries a ~9.3 solar-mass dark companion in integer appointments. Each study runs the same recipe the eclipse proved out — a forcing with a computable clock and track, a public raw archive, a sealed adversary that mimics magnitude but not shape, and a standing future event where a frozen exact-integer law must resolve in the open.

Start at the Shear Studies — Program index · read the Shear Studies White Paper.